Targeted flight rerouting for contrail avoidance relies on computational models to predict which avoidance actions would provide the most warming reduction. In the absence of empirical data for model validation, we propose using intercomparison to evaluate whether the lifetime total extinction of contrails (a proxy for climate impact) – as predicted by a high-fidelity contrail model – can be robustly predicted by lower-fidelity models when simulating contrails individually without feedbacks with a coupled weather model. We compare three contrail-cirrus models in the diffusion regime: CoCiP (a 0D plume model), APCEMM (a 2D simplified model), and EULAG-LCM (a large eddy simulation model applied on a 2D grid) under prescribed meteorological conditions. We identify that differential sedimentation and turbulence might be necessary to simulate a contrail in a physically consistent way. Since monodisperse 0D models are unable to represent differential sedimentation, the CoCiP and EULAG-LCM results have a low Pearson correlation coefficient of 0.50, compared to 0.93 between APCEMM and EULAG-LCM. The 0D and monodisperse nature of CoCiP affects its response by varying amounts in different regions of the parameter space. The presence of an explicit turbulence solver in EULAG-LCM enhances the ice crystal loss and increases the total ice mass, leading to a contrail with a higher settling velocity than in APCEMM, which parametrises turbulence. APCEMM contrails hence last longer and are larger. Hence, APCEMM overpredicts the lifetime total extinction by ~30 % relative to EULAG-LCM, demonstrating that a strong correlation between models does not imply agreement in the predicted plume-scale properties.